From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Chris Friesen" Subject: question about niu linux driver behaviour Date: Wed, 16 Sep 2009 15:13:00 -0600 Message-ID: <4AB154DC.7070305@nortel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit To: netdev@vger.kernel.org, "David S. Miller" , Santwona.Behera@sun.com, Matheos Worku Return-path: Received: from zcars04e.nortel.com ([47.129.242.56]:59552 "EHLO zcars04e.nortel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753583AbZIPVOQ (ORCPT ); Wed, 16 Sep 2009 17:14:16 -0400 Sender: netdev-owner@vger.kernel.org List-ID: Hi all, I've got a Sun 3320 ATCA "alonso" board with Neptune network devices. I'm running basically 2.6.27.8, and I'm seeing some odd behaviour in the niu driver. My understanding is that there are four ports. 0 and 1 are XMAC links, while 10, and 11 are BMAC links. On our board these normally map to eth4/5/6/7. At boot time, if the other end for the XMAC links is not up, then we see messages that look like: niu 0000:08:00.0: niu: Port 0 signal bits [00000000] are not [30000000] niu 0000:08:00.0: niu: Port 0 10G/1G SERDES Link Failed niu 0000:08:00.1: niu: Port 1 signal bits [00000000] are not [0c000000] niu 0000:08:00.1: niu: Port 1 10G/1G SERDES Link Failed and the BMAC links come up as eth4/eth5. This seems to be due to the fact that we're checking ESR_INT_SIGNALS for ESR_INT_DET0_P0 and ESR_INT_DET0_P1, and then failing because they're not set. (The 10G initialization behaves similarly.) Is this valid behaviour? I seems fragile for successful serdes initialization to depend on the other end of the link being up at driver init time. Chris